2

for homework I'm suppose to create a multi layer perceptron artificial neural network that does classification.

I'm new to PyBrain and I'm trying to create a feed forward neural network with back propagation and after googling around, it seems like there are two ways to do it in PyBrain, use buildNetwork or FeedForwardNetwork.

Additionally, I learned in class that perceptrons have step functions and from what I've seen in the documentation there doesn't seem to be a step function option.

I can't find the difference between the two and therefore can't determine which would be better for my task.

Thank you

itsSLO
  • 359
  • 1
  • 4
  • 10

1 Answers1

2

buildNetwork is sufficient to build standard networks like an MLP and it allows for enough flexibility to set the number of hidden layers, input and output neurons and the different activation functions for hidden and output neurons. Here though it will create a network with all neurons connected to form all possible paths.

If you want more customization in terms of different activation functions for different hidden layers, not all layers to be connected but selective connections, or constructing recurrent networks like those of Elman network then FeedForwardNetwork will be required as buildNetwork doesn't let you do all the necessary adjustments in the structure.

For simple perceptrons and multi-layered perceptrons there won't be any difference between the two. For you task buildNetwork should be sufficient.

Gaurav
  • 1,597
  • 2
  • 14
  • 31